Presenter/Author Information

T. Räsänen
Mikko Kolehmainen

Keywords

positioning of masses of people, online monitoring, predicting, self-organizing map, multilayer perceptron

Start Date

1-7-2008 12:00 AM

Description

The number of people in a certain place in desired time period is one of the main questions in many monitoring and management applications. Such information is needed also in tourism which has been one of most growing business areas in recent years. Increasing volume forces development of new practises for the sustainable and efficient management of recreational areas and tourist centres. Regional visitor monitoring provides general tools for managers and decisionmakers to handle multidimensional growth and the development of tourism business. Our study aims to develop predictive approach for solving continuously regional visitor attendance levels using inferred data and computationally intelligent methods. In opposite to other known visitor monitoring systems, we used mobile telecommunications network to provide data for creation of estimate for number of people in certain region. Furthermore, regional weather conditions, traffic density, restaurant sales and the use of accommodation facilities were coupled together with mobile telecommunication event data. The Selforganizing Map (SOM) was used to integrate these variables into a combined regional attendance index, and the multilayer perceptron (MLP) was used to create the shortterm predictions of visitor attendance levels. Finally the proposed continuous modelling system consisted online data gathering, server based modelling core and webbased user interface for information sharing. The system was tested and validated using real data gathered from the recreational area of Tahko. The regional visitor attendance level model and predictions were validated against expert opinions and regional freshwater consumption data. The results showed in general that the method is suitable for describing a real regional situation and seasonal variations in visitor attendance levels. Moreover, the results indicated that mobile telecommunication data improves predictions of daily visitors. Nevertheless, feedback from the possible endusers showed that presented method has potential in applications also in many other fields than tourism.

Share

COinS
 
Jul 1st, 12:00 AM

Neural Network based method for predicting regional visitor attendance levels in recreational areas

The number of people in a certain place in desired time period is one of the main questions in many monitoring and management applications. Such information is needed also in tourism which has been one of most growing business areas in recent years. Increasing volume forces development of new practises for the sustainable and efficient management of recreational areas and tourist centres. Regional visitor monitoring provides general tools for managers and decisionmakers to handle multidimensional growth and the development of tourism business. Our study aims to develop predictive approach for solving continuously regional visitor attendance levels using inferred data and computationally intelligent methods. In opposite to other known visitor monitoring systems, we used mobile telecommunications network to provide data for creation of estimate for number of people in certain region. Furthermore, regional weather conditions, traffic density, restaurant sales and the use of accommodation facilities were coupled together with mobile telecommunication event data. The Selforganizing Map (SOM) was used to integrate these variables into a combined regional attendance index, and the multilayer perceptron (MLP) was used to create the shortterm predictions of visitor attendance levels. Finally the proposed continuous modelling system consisted online data gathering, server based modelling core and webbased user interface for information sharing. The system was tested and validated using real data gathered from the recreational area of Tahko. The regional visitor attendance level model and predictions were validated against expert opinions and regional freshwater consumption data. The results showed in general that the method is suitable for describing a real regional situation and seasonal variations in visitor attendance levels. Moreover, the results indicated that mobile telecommunication data improves predictions of daily visitors. Nevertheless, feedback from the possible endusers showed that presented method has potential in applications also in many other fields than tourism.